Software NewsAnalysis/9 min read

Modern applications are mostly other people’s code. That fact built the open-source economy, and it also built the attack surface that defenders now spend their budgets on. In 2026, supply chain security has moved from conference talk to contract clause.

Key takeaways
  • Most of what ships in a typical application is third-party code, so the build and dependency pipeline is part of your attack surface.
  • An SBOM is an inventory, not a defense: its value comes from being generated automatically, kept current and wired into vulnerability response.
  • Signing and provenance let consumers verify who built an artifact and how, which blocks whole categories of tampering.
  • The EU Cyber Resilience Act and US procurement rules are turning good practice into legal and commercial requirements.
  • A short list of cheap controls, including lockfiles, pinned actions and least-privilege CI tokens, closes many common gaps.

01How supply chain risk became a boardroom topic

Three incidents explain most of the shift. In 2020, attackers compromised the build system of the SolarWinds Orion platform and slipped a backdoor into signed updates delivered to thousands of customers. In December 2021, the Log4Shell vulnerability in the ubiquitous Log4j logging library showed how many organizations had no idea where that library lived inside their own products. In 2024, the xz Utils backdoor demonstrated a patient social-engineering campaign against a lone open-source maintainer, caught only because an engineer noticed a half-second delay in SSH logins.

Each case exposed a different weakness. SolarWinds was a build integrity problem. Log4Shell was an inventory problem. The xz episode was a trust and sustainability problem. Together they convinced regulators, insurers and customers that securing your own code is not enough, and that the question what is in this software, and how was it made has to have an answer.

You cannot patch what you cannot find, and you cannot trust what you cannot verify.

02SBOMs: from checkbox to working tool

A software bill of materials is a machine-readable list of the components inside a piece of software. The two dominant formats, SPDX and CycloneDX, describe packages, versions, licenses and relationships. US Executive Order 14028 in 2021 pushed SBOMs into federal procurement, and similar expectations have spread through enterprise contracts and sector regulators.

The early experience was disappointing for many teams. SBOMs were generated once for an audit, filed away, and never consulted. Quality varied widely, with missing transitive dependencies, vague version strings and no connection to what actually ran in production. The mature approach looks different. Generate the SBOM automatically in the build, from the same lockfiles and container layers that define the artifact. Store it alongside the release. Feed it into a vulnerability database query so that when a new advisory lands, you can list every affected product in minutes.

A companion format has become important here. VEX, the Vulnerability Exploitability eXchange, lets a vendor state that a given vulnerability exists in a component but is not exploitable in this product. That cuts the noise that otherwise makes SBOM-driven scanning exhausting, because customers can see which findings truly need action.

03Signing and provenance: proving where artifacts came from

An inventory tells you what is inside. Provenance tells you where it came from. The idea is to attach a verifiable statement to each build artifact saying which source commit, which builder and which steps produced it, and to sign that statement so tampering is detectable.

Open tooling has made this far more practical than it was a few years ago. The Sigstore project offers keyless signing, in which a short-lived certificate is bound to an identity from your CI system, so teams no longer have to guard long-lived private keys. The SLSA framework, short for Supply-chain Levels for Software Artifacts, defines graduated levels of build integrity, from basic provenance to hardened, isolated builders. Attestation formats from the in-toto project let multiple parties sign statements about the same artifact, such as a build, a vulnerability scan and a review.

Signing only helps if someone checks. The decisive step is enforcement at the point of use: a Kubernetes admission policy that rejects unsigned images, a package manager that verifies attestations, or an internal registry that refuses artifacts without provenance. Without verification, signatures are decoration.

ControlWhat it mainly stopsTypical effortWhere to start
Lockfiles and pinned versionsSurprise upgrades, dependency confusionLowCommit lockfiles, forbid floating versions in production builds
Pinned CI actions by commit hashHijacked third-party build stepsLowReplace tag references in workflow files
Least-privilege CI tokensLateral movement from a compromised jobLow to mediumDefault tokens to read-only, grant writes per job
Automated SBOM generationSlow incident response, blind spotsMediumEmit an SBOM in every release pipeline
Keyless signing and provenanceTampered or counterfeit artifactsMediumSign container images and release binaries
Admission-time verificationDeploying unverified artifactsMedium to highEnforce in one non-critical cluster first

04The regulatory clock is ticking

Compliance is now a major driver. In the European Union, the Cyber Resilience Act entered into force in December 2024 and sets security requirements for products with digital elements sold in the EU market. Its obligations phase in: manufacturer reporting duties for actively exploited vulnerabilities and severe incidents begin in September 2026, and the main conformity requirements apply from December 2027. Those requirements include handling vulnerabilities throughout a support period and maintaining documentation that effectively includes an SBOM.

In the United States, federal procurement has relied on secure software development attestations built around the NIST Secure Software Development Framework, and many large enterprise buyers copy those expectations into their own questionnaires. The commercial effect is the same everywhere: a vendor that cannot show how its software is built and what it contains will lose deals to one that can.

Open-source maintainers are also in the frame. The Cyber Resilience Act includes carve-outs and a steward role for non-commercial open-source projects, but companies that ship products built on open-source components are expected to exercise due diligence on them. Expect more corporate funding of maintainers, more security audits of critical packages, and more pressure to contribute fixes upstream rather than patching privately.

05Dependency hygiene that actually works

Most organizations get more risk reduction from boring practices than from advanced tooling. Start with ownership: every repository should have a named team responsible for its dependencies. Automate updates with a bot, but add a short cooling-off period before adopting brand-new releases, since many malicious package versions are caught and pulled within days of publication.

Use an internal package proxy so that builds fetch from a controlled mirror, and so that you can block a bad version everywhere at once. Prefer dependencies with active maintainers, a visible security policy and a track record of timely fixes, and treat abandoned packages as technical debt with a deadline. For critical components, review the changelog and the diff of upgrades rather than trusting version numbers.

Finally, reduce what you depend on. Every package is a promise to keep reading its advisories. Replacing a large library you use for one function with a few lines of your own code, or with the standard library, is often the cheapest vulnerability fix available.

06Hardening the build system itself

Attackers have learned that build and release infrastructure often has weaker protection than production. Workflows with write access to repositories, registries and cloud accounts are lucrative targets. Treat CI as production: restrict who can modify pipeline definitions, require review for changes to them, isolate runners between projects, and never expose secrets to builds triggered by untrusted pull requests.

Ephemeral, hermetic builds are the gold standard. A builder that starts clean, fetches only declared inputs, produces an artifact and disappears leaves no persistent foothold, and it makes the output reproducible enough that independent parties can rebuild and compare. Reproducible builds remain difficult for many ecosystems, but even partial progress shrinks the room an attacker has to hide.

07A twelve-month roadmap

1
Quarter one: see what you have
Inventory repositories and owners, turn on dependency scanning, and emit an SBOM from the build of your most important product.
2
Quarter two: close the cheap gaps
Pin CI actions, make default tokens read-only, enforce lockfiles, and put a proxy in front of public registries.
3
Quarter three: sign and verify
Adopt keyless signing for release artifacts, publish provenance, and enforce verification in one environment.
4
Quarter four: connect to response
Link SBOMs to vulnerability feeds, practice a Log4Shell-style drill, and document your vulnerability disclosure and reporting process for regulators.

08Frequently asked questions

Is an SBOM required for every software company?

Not universally, but requirements are spreading through government procurement, sector rules and the EU Cyber Resilience Act. Even where it is optional, an accurate SBOM shortens incident response dramatically, so most teams benefit from generating one.

Does signing prevent malicious open-source packages?

No. Signing proves origin and integrity, not intent. A malicious maintainer can sign malicious code. Signing is one layer alongside review, scanning, cooling-off periods and behavior analysis.

What is the difference between SPDX and CycloneDX?

Both describe software components in machine-readable form. SPDX began with license compliance and is an ISO standard, while CycloneDX grew out of application security. Many tools can produce or convert either, so pick the one your customers and tools consume.

Where should a small team begin?

Commit lockfiles, pin CI actions, restrict CI token permissions and enable automated dependency alerts. These steps cost little and address the most common real-world weaknesses.

The bottom line
Supply chain security rewards unglamorous work: knowing what you ship, controlling how it is built, and verifying what you deploy. The incidents of the past few years were not failures of exotic technology. They were failures of visibility and trust. Teams that invest in both now will be ready for the audits, the customer questionnaires and, eventually, the next big advisory.

Leave a Reply

Your email address will not be published. Required fields are marked *